A. Pomerance, D. Canaday, N. Charlot, D. Gauthier
Potomac Research LLC, United States
Keywords: physically uncloneable functions, field programmable gate arrays, cybersecurityPhysically Uncloneable Functions (PUFs) have been proposed as a cybersecurity primitive for tamper-proofing, device fingerprinting, and intellectual property protection. PUFs work by exploiting small manufacturing variances so that different devices produce a unique, unpredictable response when queried with a specific challenge. Despite their promise, there are two key problems with PUFs that limit their adoption in practice. Many designs appear to be vulnerable to machine learning attacks that can successfully predict responses, thus negating the security benefit, and the reliability of responses can be sensitive to temperature and voltage variations so that the fingerprint depends on the environment. Here we present a novel PUF design based on Hybrid Boolean Networks, the HBN-PUF, that appears to address these shortcomings in a compact field programmable gate array (FPGA) circuit. The HBN-PUF uniquely produces a multi-bit responses. As a result, HBN-PUFs have orders of magnitude more entropy and are likely more resistant to machine learning attacks compared to other designs. Furthermore, environmental testing shows that HBN-PUFs are relatively robust to temperature variations without expensive temperature compensation circuitry.